🚀 go-pugleaf

/ gnu.misc.discuss / jhummel@wave.ics.uci.edu (Joseph Edward Hummel)/Jul 20'91/
> I think in general it is an idea long overdue [...]
> We all know computers are being embedded in more and more products, and
> how do we know these computers are being programmed in a competent
> manner?

We make people and companies responsible for what they produce.  I.e., none
of those product licenses that say:

	Yeah, we know that in our ads we shouted that this program does
	this and that, but this here fine print is here to tell you that
	the only thing we promise is that a goat didn't chew on the
	enclosed storage device on its way from us to you.  By using our
	product, about which, you must note, we make no promises that it
	works at all, or does what you think it's supposed to do, or does
	what we advertised it to do, you agree not to sue us.

Holding people responsible for their products (by making such clauses null
and void) would do much better for consumer protection than a licensing
procedure which, judging by a quick overview of the current job market,
would classify you as "professional" if you can pass a test on CICS, SAS
and COBOL.

> How about the recent X-ray machine, programmed by an individual with a
> 2-year degree, that killed 3 or so people?

How about a program that solves the quadratic equation, that a sufficiently
curious junior high school student is quite capable of writing?

I assure you that there are plenty of CS Ph.D.'s that would kill Superman
if they were allowed to program an X-ray machine.

> And what about companies in
> financial trouble trying to get a product out the door as cheaply as
> possible?

Is the solution to price them out of the market by legislating minimum salary
requirements???

> There are no minimal standards that these companies nor
> individuals have to follow

Do we require auto makers to hire P.E.'s exclusively?  Or do we place
minimal safety requirements on the PRODUCTS they make?

> -- I would hope there are some minimal standards
> we could all agree on:

Well, let's see...

> you must have a design document

Nonsense.  Whether or not a design document is beneficial depends very much
on the problem being solved.  If it is beneficial, how it is developed and
what form it takes depends even more on the problem being solved.  I
certainly do NOT agree that there must be a "design document" just so some
bureaucrat can stamp it "Approved".

> you must hire programmers with a degree in CS

No way!  I know way too many excellent programmers who didn't major in CS
or didn't finish their formal degree, and many more horrible programmers
who have a CS degree -- at all levels!  In fact, I find that most people
who pride themselves on being CS "professionals" would barely pass the
beginner CS courses that I teach.

> (or have them take a language-independent
> exam, along the lines of the GRE exam for CS),

Oh yeah.  COBOL programmers sure do need regular expression and finite
automata theory.  And grammars, yes, certainly.  How could they ever
PERFORM a SORTing paragraph without knowing about context-free grammars.
And how could they ever invert your social security number without knowing
linear algebra!  That would most certainly create unsafe accounting
systems...

> and you must test your system and document your test cases.

Wonderful.  The first person to develop a practical way to derive provably
correct systems will get a Nobel Prize, but will be unable to pass
certification.

I don't care what you did to make your system safe and reliable, as long as
it is.  I don't care that you tested your system and wrote up what you did
(are you going to have a gauge for thoroughness of testing?).

I don't care what you studied.  I don't want you to be certified or
certifiable.  I want your product to be good.

> Or perhaps a suite of tests, one for Business programming, one for
> scientific programming, one for systems programming, one for embedded systems
> programming, etc.  These tests should be easy for anyone already competent,
> so what is everyone (e.g. AT&T) worried about?

Because we don't know what "basics" are in any of these fields.  Sure, we
can figure out what the majority out there is doing, but could you consider
the possibility that most of the rest of us may not be doing it the same
way for reasons other than incompetence?

Any codification of practices at this stage of computer science will be
much more POLITICAL than technical.

Test programs, not programmers.

Jacob
--
Jacob Gore		Jacob@Gore.Com			boulder!gore!jacob